Helpful to verify reports of your own downtime, or to double check a downed website you are trying to access.
One of a set of tools we are providing to everyone as a way of saying thank you for being a part of the community.
ZFS Nexenta & ESXi: some more practical questions for optimizing hardware config
Hi,
I have a HP Proliant NL40 configured with Nexenta.
Current config:
*16GB RAM
*3 SATA disks 250 GB RAIDZ with 2 mirrors for log and cache (Kingston entrerprise ssd's of 120 GB)
*OS on 60GB mirrored drive OCZ Vertex
Please advise on this (what I could change to current config):
--------------------------
1.buy two laptop disks for os and install os on it (mirrored)
2.put some extra sata disks (4) in, I have 5x 2TB Seagate barracuda in my Synology NAS (RAID5)
=> I thought of using the same Barracuda 2TB x 4, then I could swap in case of issues (and don't need to go to store the same day a disk crashes)? If not these disks, which would you recommend?
3. I would only use the Nexenta for ESXi, so I don't need that much storage then (only 2 or 3TB max) UNLESS I would also put my data on it (documents), would you recommend or not? If not, then maybe smaller disks would optimize performance?
4.use the 3 SSD's of my ESX-servers (60 GB OCZ) for log (mirror) and
=> cache 1 disk? => I could order an extra one if it would be worth it
5.then I have 4 ssd's of 120 GB (Kingston), here I wonder what the best is I could do: use 2 of them as swap file location + host cache for my two ESX-es, then I still have 2 ssd's of 120 GB, one could fit in a laptop I barely use, the other ...
I guess there is no use adding 2 SSD's of 120 GB per ESX.
6.What technology/ies have you running for storage NFS or ISCSI?
7.Do you have other things installed on ZFS system, f.e. ftp, monitoring ...?
I had comment already the SSD's of 120GB are overkill, so that's why I'm looking for optimizing the use of my config, hence the question. SSD's of 60 GB (OCZ) would better fit for log/cache then the 120GB Kingston, but then again they are enterprise version ssd's.
Please advise.
J.
I have a HP Proliant NL40 configured with Nexenta.
Current config:
*16GB RAM
*3 SATA disks 250 GB RAIDZ with 2 mirrors for log and cache (Kingston entrerprise ssd's of 120 GB)
*OS on 60GB mirrored drive OCZ Vertex
Please advise on this (what I could change to current config):
--------------------------
1.buy two laptop disks for os and install os on it (mirrored)
2.put some extra sata disks (4) in, I have 5x 2TB Seagate barracuda in my Synology NAS (RAID5)
=> I thought of using the same Barracuda 2TB x 4, then I could swap in case of issues (and don't need to go to store the same day a disk crashes)? If not these disks, which would you recommend?
3. I would only use the Nexenta for ESXi, so I don't need that much storage then (only 2 or 3TB max) UNLESS I would also put my data on it (documents), would you recommend or not? If not, then maybe smaller disks would optimize performance?
4.use the 3 SSD's of my ESX-servers (60 GB OCZ) for log (mirror) and
=> cache 1 disk? => I could order an extra one if it would be worth it
5.then I have 4 ssd's of 120 GB (Kingston), here I wonder what the best is I could do: use 2 of them as swap file location + host cache for my two ESX-es, then I still have 2 ssd's of 120 GB, one could fit in a laptop I barely use, the other ...
I guess there is no use adding 2 SSD's of 120 GB per ESX.
6.What technology/ies have you running for storage NFS or ISCSI?
7.Do you have other things installed on ZFS system, f.e. ftp, monitoring ...?
I had comment already the SSD's of 120GB are overkill, so that's why I'm looking for optimizing the use of my config, hence the question. SSD's of 60 GB (OCZ) would better fit for log/cache then the 120GB Kingston, but then again they are enterprise version ssd's.
Please advise.
J.
Who is Participating?
No need to mirror cache. You are protected against data loss by zfs if a non-mirrored cache drive fails. If you need more cache, just add a non-mirrored SSD.
1. laptop disks are just fine for the o/s. That is what I have in my home server.
2. With cheap desktop drives, go RAIDZ2, not RAIDZ1. You do NOT want to sweat 100% data loss when it takes 24 hours or more of intense I/O doing a rebuild ... after you discover you had a drive failure.
Get enterprise class SATA if budget allows. Benefits go way beyond 24x7 capability.
If you want to use those extra SSDs, create a separate mirrored pool for them, and put data files there like database index files, scratch table space, and anything else that requires high-I/O. NO need to put a ZIL or cache on SSD-based mirrored pool.
1. laptop disks are just fine for the o/s. That is what I have in my home server.
2. With cheap desktop drives, go RAIDZ2, not RAIDZ1. You do NOT want to sweat 100% data loss when it takes 24 hours or more of intense I/O doing a rebuild ... after you discover you had a drive failure.
Get enterprise class SATA if budget allows. Benefits go way beyond 24x7 capability.
If you want to use those extra SSDs, create a separate mirrored pool for them, and put data files there like database index files, scratch table space, and anything else that requires high-I/O. NO need to put a ZIL or cache on SSD-based mirrored pool.
@hanccocka thanks!
But don't understand this:
Number 5. why would I "configure as ssd", it is already ssd, you mean when it wouldn't me recoginzed as ssd?
Would you advise to use full 120 GB as host cache AND swap file location?
@dlethe also thanks for your input!
raidZ2 is ... "normal raid10"? so you would suggest RAID10 over 4 disks?
Use those extra ssd's, how can I? All my drives are full ... 2 for os, 2 for log (mirror), 1 for cache ... then there is only 1 ssd over in my sharkoon?
>NO need to put a ZIL or cache on SSD-based mirrored pool.
That's what hanccocka adviced ...? Mirrored log on ssd, cache also on ssd (but only 1 ssd needed). So you would only use 1 ssd for log, 1 for cache, 2 for os, then yes I would have the possibility of making a mirrored volume for data (2x ssd 120GB).
But don't understand this:
Number 5. why would I "configure as ssd", it is already ssd, you mean when it wouldn't me recoginzed as ssd?
Would you advise to use full 120 GB as host cache AND swap file location?
@dlethe also thanks for your input!
raidZ2 is ... "normal raid10"? so you would suggest RAID10 over 4 disks?
Use those extra ssd's, how can I? All my drives are full ... 2 for os, 2 for log (mirror), 1 for cache ... then there is only 1 ssd over in my sharkoon?
>NO need to put a ZIL or cache on SSD-based mirrored pool.
That's what hanccocka adviced ...? Mirrored log on ssd, cache also on ssd (but only 1 ssd needed). So you would only use 1 ssd for log, 1 for cache, 2 for os, then yes I would have the possibility of making a mirrored volume for data (2x ssd 120GB).
RAIDZ2 is best compared to RAID6. A 4-disk RAIDZ2 has 2 parity disks. RAID10 has 1 parity drive for 50% of failure scenarios, and 2 parity drives for 50% failure scenarios. As one has to plan for worst-case scenario, then RAID10 only has 1 parity disk.
if it's not recognized as SSD!
120GB is large for host cache, on such a small server.
as you have such a large SSD, you could just move all the Swap files there, and not use Host Cache.
but you do not have to allocate it all, you could also use as a data store.
You ALWAYS need to MIRROR ZIL (LOG), when used with SATA disks!
120GB is large for host cache, on such a small server.
as you have such a large SSD, you could just move all the Swap files there, and not use Host Cache.
but you do not have to allocate it all, you could also use as a data store.
You ALWAYS need to MIRROR ZIL (LOG), when used with SATA disks!
You always need to mirror ZIL, period .. unless you have purchased hardware that does not run in this universe and does not follow the rules of entropy. 100% of hardware fails, eventually.
(But ZFS does not require you to mirror ZIL, you need to mirror it if you want to decrease risk of data loss due to the statistical certainty of component failure).
(But ZFS does not require you to mirror ZIL, you need to mirror it if you want to decrease risk of data loss due to the statistical certainty of component failure).
Note: I have two different older laptops (some years) of which disk drives are different (in type as well as in size, think one is about 200 GB other 500 GB), could I use them for OS mirror ... or do they need to be exactly the same?
you can use them, just make sure the parititons are the same size. (but Nexenta will complain at installation, if it cannot use them!)
Note: 2TB disks for NL 40 (Microserver) still seem very expensive (4 * 100 Euro's), any other suggestion ... budgetwise? Would like to stick to 2TB to be able to put them in synology in case of disk failure.
Note: if you would like new ticket for this, no problem
Note: if you would like new ticket for this, no problem
Final questions: the faster I make these SATA's (I need sata's not sas drives, right), the faster ZFS will be?
Or does it really not matter, thanks to the cache/logs on ssd?
Or does it really not matter, thanks to the cache/logs on ssd?
Question has a verified solution.
Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.
Have a better answer? Share it in a comment.
All Courses
From novice to tech pro — start learning today.
-
Course of the Month5 days, left to enrollMonitoring and Operating a Private Cloud with System Center 2012 R2 273 lessons
2. Low Power, 5,400rpm disks are fine, because you are using SSD cache
3. we use for CIFs (Windows Shares)
4. Memory in the Server is also used for ARC cache (so more than 1 SSD is probably overkill!)
5. Yes, see my EE Articles
HOW TO: Tag and Configure a storage device as a Solid State Disk (SSD) in VMware vSphere 5.0 or 5.1 (ESXi 5.0 or ESXi 5.1)
HOW TO: Configure Host Cache Configuration for your VMware vSphere 5.0 or 5.1 ESXi (5.0 or 5.1) Host Server
6. We use both NFS and iSCSI
60GB or smaller SSDs are better for logs and cache.
7. No we do not have any other monitoring, these are SANs.